Evaluation of an Automated Smartphone-based Digital Auscultation Application for Detecting Abnormal Heart Sounds Using Deep Learning Techniques

Overview

This pilot study is to investigate the feasibility of obtaining medical grade audio phonocardiogram (PCG) recordings using a smartphone-based auscultation device in the first step. The ability to determine Valvular Heart Disease (VHD) (i.e., presence or absence of cardiac murmurs) using novel handheld CAA-devices shall be analyzed and first data on a smartphone-based auscultation in a hospital setting shall be collected. In further studies, the data provided from this study can be used to investigate the potential diagnostic use of such devices in the ambulatory and stationary care scenarios.

Cardiac auscultation is considered to be highly subjective with substantial varying sensitivities and specificities in regard of the practitioners' expertise. Computer-assisted auscultation (CAA) aims to provide increased objectivity. CAA makes auscultation procedure less operator-dependent, approximate inter-examiner differences and may reduce uncertainties in the course of the examination. With the introduction of modern Machine Learning software libraries and ever-growing computational resources CAA has advanced significantly and is now able to classify heart sounds and murmurs into normal and abnormal, using complex spectro-temporal signal processing techniques and neural network pathways. CAA has simultaneously made the shift from the deployment on computers to consumer smartphones. A benefit of CAA can be expected from the smartphone alone in terms of cost, application range, the clinical validity of such algorithms should now be measured in this pilot study. This pilot study is to investigate the feasibility of obtaining medical grade audio phonocardiogram (PCG) recordings using a smartphone-based auscultation device in the first step. In further studies, the data provided from this study can be used to investigate the potential diagnostic use of such devices in the ambulatory and stationary care scenarios.

Conditions